Follow Slashdot stories on Twitter

 



Forgot your password?
typodupeerror

Comment Re:What's with the Mechanical Turk-based "research (Score 2, Insightful) 293

Pretty much all these quantitative sociology studies turn out to be questionable. Between populations of convenience and these hokey numerical personality tests, the results don't inspire that much confidence.

Even though in my day job I do statistics, when it comes to social science I often find qualitative anthropology/ethnography-style research, where researchers actually get out there in communities, try to understand them, and talk to people, quite a bit more informative. Especially for preliminary understanding where it's not often even clear what a phenomenon's broad characteristics are, and therefore difficult to design an intelligent quantitative study with useful metrics.

Alas, this kind of stuff gets more citations and press, because sampling 5000 people and rating them on a 0.0 to 10.0 personality scale using a questionnaire seems superficially more scientific... 'cause it's got numbers.

Comment Linux is dying (Score 2) 53

It is now official. The Università degli studi di Milano has confirmed: Linux is dying.

One more crippling bombshell hit the already beleaguered Linux community when UNIMI confirmed that Linux's flagship domain, kernel.org, fell to a shocking #1797 in the Common Crawl rankings. You don't need to be the Amazing Kreskin to predict Linux's future. Its domain now ranks just behind Excite.com, the now-irrelevant search engine from the 1990s, which edges it out at #1796.

The glaring gap between Linux's ranking and the rankings of those in the vibrant, enterprise-ready world is in itself embarrassing enough: Apple #8, Microsoft #17, even Oracle #248. But what seals the coffin is that Linux has fallen behind even the notoriously moribund FreeBSD operating system in these industry-leading metrics, trailing it by nearly one thousand, five hundred positions.

Comment Re:gmpg.org? (Score 1) 53

There's an extension to the <link rel> tag that overloads it by, instead of linking to actual related data (as the tag was intended to do), treats the target of the link as defining a schema / data format, when rel="profile". The URL is then essentially a globally unique key for the data format; parsers that recognize the format will see the key and know how to parse some other information on the page. gmpg.org is the host of one of the early ones, XFN, which is linked in default Wordpress installs.

Comment Re:Edit, but disclose (Score 2) 112

That's also more or less the actual Wikipedia policy, despite what the summary implies. There is no prohibition on people who are paid money also editing Wikipedia. In fact Wikipedia actively encourages it in some cases, such as trying to recruit more museum and library staff to contribute to Wikipedia.

What is generally prohibited is: 1) taking money to write promotional articles or "clean" articles on behalf of their subject, like Wiki-PR was doing; or 2) making edits on an article where you have a clear COI, and failing to disclose your relationship to the subject of the article. In practice even #2 only really matters if they're biased edits that would cause someone to raise an eyebrow.

Comment Re:Linux vs. Hurd/xBSD (Score 3, Interesting) 163

I agree it's basically a confluence of circumstances. Fwiw, while GNU's kernel project was pretty unsuccessful, I do think their more general project of trying to put together all the parts of a free Unix-like was quite useful, and one part of the circumstantial confluence. With BSD tied up in licensing issues at the time, Linus was able to basically grab the GNU compiler, libc, userland, etc. and make a working system. GNU's efforts were less essential to the BSDs after the lawsuit was resolved, but still fairly important in the early years to get something up and running: the lawsuit resolution resulted in ripping out the AT&T-licensed code from BSD, a bunch of which was replaced by GNU utilities as drop-in replacements. These have since been re-replaced in most of the BSDs ('grep' was one of the last GNU utilities to be phased out), but served as a pretty useful 20-year stopgap. And of course GCC had replaced the traditional CC much earlier (GCC appears in 4.3BSD).

One missing bit of this soup that's a real shame, imo: The very late open-sourcing of Plan9 led to a bunch of good stuff that could've been pulled in being ignored. If at least parts of Plan9 had been available in the early '90s when this GNU/BSD/Linux code was coalescing into free operating systems, Plan9's code could've contributed usefully.

Comment Re:Micro Kernel, Failed Computer Science Pipe drea (Score 2) 163

An interesting historical tidbit about QNX is that it was started more or less on the basis of a textbook implementation of a microkernel with real-time features. In the literal sense that the company's co-founders did a class project where they implemented a basic realtime microkernel in an OS class, wondered why there wasn't something similar in the marketplace, and founded a company to sell it.

Comment don't think this really works on phones yet (Score 5, Interesting) 151

In an age where our phones have accelerometers and compasses, it's amazing your car is still trying to catch up, right?

Actually I think it's the opposite, it's only being in a car that makes dead reckoning with any kind of accuracy feasible. A car is a reasonably large and stable platform, which already has good speed information, and can have some accelerometer-type information added relatively easily. A smartphone does have an accelerometer, but the data is far too noisy to do reasonable dead reckoning, because in addition to the macro movements (someone walking, biking, or driving down a street) there area bunch of micro-movements that produce high local acceleration (putting the phone in/out of pockets, taking steps while the phone's in your pocket, etc.). It makes for a much more complex dead-reckoning problem, because instead of just tracking broad movements (car goes 10m this way) you have to resolve a ton of tiny movements (phone was moved 0.3m into pocket, then rapidly accelerated 0.1m left due to owner being jostled on the subway, etc.), which tend to pile too much accelerometer noise on top of the broader movements that you really want to track.

In short, taking a known starting position and keeping it updated via accelerometer data is a lot easier if your accelerometer is on a car, vs. on a handheld device.

Slashdot Top Deals

If I have not seen so far it is because I stood in giant's footsteps.

Working...